Repository navigation
Evaluate LLM provider costs and identify free/local options #27
Description
Before building multi-model support, understand the full cost picture and ensure a free path exists.
Scope
- Ollama (local): Completely free. What models run well on typical hardware? (Llama 3, Mistral, Phi-3, Gemma). What are the quality trade-offs vs. hosted models for financial reasoning tasks?
- Claude (Anthropic): Pricing per 1M tokens for Haiku vs. Sonnet vs. Opus. What's the realistic monthly cost for a personal finance app at typical usage?
- OpenAI: GPT-4o-mini vs. GPT-4o pricing. Compare to Claude.
- Free tiers: Any hosted models with free tiers sufficient for personal use? (Groq, Together AI, Google Gemini free tier)
- Decision output: MVP should default to Ollama (free, local, privacy-preserving). Hosted models as opt-in with API key. Document the recommended default model per provider.
Activity
docs: initialize GSD planning — requirements and roadmap aligned to G…